Obesity risk in middle-aged women linked to air pollution in new study
Obesity risk in middle-aged women linked to air pollution in new study
Obesity is a huge issue, and its affected population has tripled over the past few decades (1). The medical issues that can arise from obesity are very severe, including: type 2 diabetes, heart disease, some cancer, and even death (2).
It is known that pollutants can lead to inflammation, immunosuppression, oxidative stress, and trigger mutations within human cells (13). There has been a focus on the effects of these pollutants on respiratory tissue, and some focus toward adipose tissue.
It should be noted that exposure to pollutants like NO2 and O3, as well as particulate matter seen in aerosol emissions like PM2.5, can lead to oxidative stress on cells; as a result, several obesity risk factors present themselves: adipose tissue inflammation, hypothalamic-adrenal axis dysfunction (hormonal issues), and brown adipose tissue dysfunction (3-8). These are all mentioned in the article's background as well.
There has been a lot of research done on this as it relates to childhood obesity (7). However, the focus of these studies has been on BMI and not DXA scanning, which reveals more information about body composition and can separate lean mass from body fat, which BMI does not do. Further, DXA is a very low-dose X-ray that passes over you for 5–15 minutes and tells you your bone mineral density and body composition regarding muscle and fat.
Moreover, it should be noted that the midlife is an important stage for body composition, as this is the stage in which the body is seen to have a particular vulnerability to obesity. Gains in fat mass and loss of lean muscle mass are seen to occur during this stage of life (10).
Regarding the study (SWAN), the Study of Women's Health Across the Nation was used. SWAN is an ongoing cohort study of midlife women, which had its population trimmed and adjusted to be used for this 8-year analysis (9).
The paper 'Longitudinal Associations of Air Pollution With Body Size and Composition in Midlife Women: The Study of Women's Health Across the Nation' by Xin Wang et al. in Diabetes Care conducts a longitudinal study (a study following the same people over a period of time) following 1,654 participants of a variety of races (Chinese, White, Black, and Japanese) to determine whether air pollution (NO2, O3, and PM2.5) was correlated with body composition. Air pollution exposure was estimated by linking the residential addresses of the participants with a 1 km^2 resolution pollution estimation. From yearly visits over 8 years, from 2000 to 2008, the participants had their body size and composition measured.
The first step within the study was to get the population to be analyzed. SWAN had 3,302 women enrolled in it; to be included in this initial sample, you have to be aged 42–52, have an intact uterus, have at least one ovary, have at least one menstrual cycle in the past 3 months, and have not used hormone therapy for at least 3 months prior. Then, from a series of major urban sites, a population of White women and a particular ethnic group was chosen from each site: [Black women: Boston, Pittsburgh, Southeast Michigan, Chicago], [Hispanic women: Newark], [Chinese women: Oakland], and Japanese women [Los Angeles]. This gave the researchers 2,452 participants, but Boston didn't have enough climate data, and Chicago and Newark didn't use DXA scanning to do body composition; as a result, the researchers were left with 1,675 participants.
The population was also asked a variety of questions about variables that could also play a role in the results that were observed, like physical activity, drinking habits, smoking habits, menopausal status, and daily calories. They also had their weight, BMI, waist circumference, fat mass, lean mass, and proportion fat mass (%) taken. The result of this initial intake can be seen below in Table 1.
The next step was to collect air pollution exposure measures for each of the participants. The measures taken were: (an annual measure of the average PM2.5 concentration, an annual average of 1-hour maximum NO2 concentration, an annual average of the 8-hour maximum of O3 daily). A 1 km^2 spatial resolution was taken, and three machine learning models were used to estimate daily ambient PM2.5, O3, and NO2 (neural network, random forest, and gradient boosting). These models had more than 100 predictor variables as well (land use, meteorologic, chemical transport, etc.). The predictive models were seen to be accurate with “10-fold cross-validated coefficient of determination.”
Further, linear mixed-effect models were used to see the association between measures of body size and composition and air pollution. The measures—lean mass, fat mass, weight, and waist circumference—were log-transformed because they were right-skewed and then back-transformed to give a percent change. There were certain variables/covariates adjusted for within the model, using “random intercepts were included to account for intra-participant correlations of repeated outcome measures.” These variables were: (non–time-varying variables: age, race, study site, and education) and (time-varying variables: follow-up time, financial hardship, smoking status, secondhand smoke, alcohol intake, physical activity, calories eaten, and menopausal status).
For physical activity in particular, the researchers incorporated a multiplicative interaction term with pollution. Explicitly, physical activity was treated as a continuous variable, and the effects of air pollution were calculated at various percentiles of exercise. The results of this can be seen below in Supplemental Table 3.
In summary, the study showed that air pollution resulted in higher adiposity and less lean mass in midlife women, even with increased exercise. A brief summary of statistics measured that had a significant P value is given below.
PM 2.5: 4.5% increase in fat mass, loss of 0.39% lean mass, and 1.1% increase in fat mass proportion.
NO2: 3.39% increase in fat mass, loss of 0.86% lean mass, and 0.94% increase in fat mass proportion.
O3: decrease in waist circumference by 0.28%, decrease in lean mass by 1%, and 0.25% increase in fat mass proportion.
The news article (11) introduces the information with a quick blurb about how 42% of adults in the U.S. have obesity and quickly talks about genetic and personal factors that lead to obesity. Then the author brings up the study. It says the demographic is mid-aged women with an average age of 49.2, which is accurate with the scientific article. The article states that exposure to pollution leads to a body fat increase of 4.5% or 2.6 pounds, which is a little oversold, as that is just the percentage with PM2.5.
The lead author of the study, Xin Wang of the University of Michigan's School of Public Health, was interviewed. The first excerpt from him goes over the link between pollution and the increase in inflammation of fat tissue as a risk factor for obesity.
The article also gets an outside opinion from Dr. Fatima Cody Stanford, an obesity medicine physician and researcher. Her opinion introduces another supporting point that is not mentioned in the scientific article, which cites a systematic review done in 2018 linking air pollution having an effect on the storage of cholesterol within the body. This systemic review is available for summary, but the study itself is hidden under a paywall. Further, she is also cited saying that obesity isn't just linked to air pollution and to be cautious.
Then the author of the news article gets another expert opinion from Dr. Mark Conroy to say that exercise is important for body composition.
The author then returns to the point of view of Wang and him, saying that this study was done on a very specific population and a very specific range of PM2.5 (12.3 µg/m^3 to 15.9 µg/m^3). He also says these results are a call for more studies to establish whether air pollution is a significant contributor to obesity.
The news article does a great job summarizing and applying the information in the study. The context provided prior to going into the study is amazing and both digestible and relatable for the average American. The study that is referenced is easily found and linked with a hyperlink to both the journal and the article. A brief synopsis of the study being longitudinal without saying longitudinal is included, which is easy to understand for the layman. Then, the results are immediately discussed, saying that fat percentage increases by "4.5% or 2.6 pounds." I do take slight issue with this. I would say this is somewhat polarizing and nonspecific. The article mentions PM2.5 elsewhere, but doesn't want to, here, where it is important to say so. This kind of gets at the polarization in reporting a potentially misleading statistic.
However, they do include interviews from the lead scientist in the study to offer some clarity. Another nitpick that I had with the article was that the study by Dr. Wang accounted for exercise, but there isn't a mention of that. Instead, the article consults a separate expert to say that exercise has a profound effect on obesity. Then the author of the news article gets another professional opinion from a doctor at Ohio State to say exercise is important. This doesn't really do the researchers at the University of Michigan justice for the mathematical measures taken within their study to accommodate the fact that exercise has an effect on body composition.
This was why I took away 3 points, but overall, the article did a good job introducing the general public to the topic, made its point about the findings, fast. It was pretty concise, and it was simple and straightforward to read.
Rahib, great job on this analysis — very informative and an interesting topic!
ReplyDeleteI really liked your review at the end, and I agree with you that the news article should not have left out the fact that the original researchers had taken other obesity factors into account in their studies. I actually really appreciated that the original study considered various other factors that might contribute to obesity, and also that it explicitly explained how "air pollution exposure" was measured in the article.
I also have two questions:
- Is there any specific reason for the cities or regions they chose for their study?
- Additionally, what might be some potential limitations of using residential addresses as an indicator for the air pollution exposure? For example, if someone spends more time at work than at home…
Thank you!
I think they chose these sites for two reasons. In the paper it said that these because they residential history available to generate exposure data. This is likely due to urban centers having more attention paid to emissions and air quality because of the amount of pollution generated from cities themselves. The sample itself was pulled from SWAN, so that may have also informed the choice for those locations. They used a kilometer radius to mitigate that to a certain degree, but this is valid concern for the study, honestly good critique.
DeleteHi Rahib! I really liked the paper you have picked and your analysis. I agree with your scoring. It is misleading of the article to not specify that the fat percentage increase statistic was for PM2.5 and to exclude the study's section about how exercise was accounted for.
ReplyDeleteI did want to comment that I liked figure s1 from the supplementary figures of the paper that I felt summarized the study design nicely. From figure s3 from the supplementary figures, I noticed that O3 and NO2 measurements seemed relatively consistent but PM2.5 measurements seem to decrease slightly. I wonder if the study mentions why or if there is any events that had happened between 2004-2008 to account for this decrease.
I am also curious if the paper went into more detail about the findings from supplementary figure, table 2 and about findings within communities, such as the predicted effect of these factors on their measurements/findings.
They used a bunch of predictive modeling methods like neural network, random forest, and gradient boosting. The paper said that these modeling methods with the 100+ variables that they included produced a fairly accurate predictive model, and there results seemed to align with what they predicted. I am unsure of what you mean with the inaccuracy from 2004-2008. Within the study, there is more variation from 2000-2003 in PM2.5 with it having a greater range. I also see that their is greater range in O3 and NO2 across all percentiles.
DeleteHi Rahib! I thought this was a very interesting blog post. I thought the research study was crafted well in the sense that it accounts for many factors that make the study more applicable. I also thought it was really interesting how they analyzed separate results for PM2.5, NO2, and O3. Additionally, Table 1 was very helpful in showing the array of lifestyle choices represented in this study. To what extent do you think this study can be generalized to populations in areas outside of the cities sampled with different pollution exposures? I agree with your scoring as the body fat increase statistic seems cherry-picked to get reactions out of readers. Although having multiple professional opinions increases the credibility of the news article, leaving out main points can do the opposite.
ReplyDeleteHi Kristen, I think it could be applied to other areas with more analysis of where wind is taking the pollutants. Then it could likely be applied to more rural areas to see how those areas are effected. In theory, the expert opinions increase the validity of the article but seeing how critical information from the study was left out decreased the validity in my eyes.
DeleteHi Rahib; I really enjoyed reading your analysis, it was definitely eye-catching from the start. An interesting exploration of a connection between two things that would seem to be unrelated. It is very impressive that the test tried to account for as many variables as possible for their participants. I do wonder if this analysis can be expanded into different age groups and if there are changes depending on a woman's fertility, as that was one of the constrictions for participants. I agree with your scoring as it does seem like the news article tried to seems to be picking and choosing with stats to keep and which to throw to best write the most impactful article, not the most transparently informational.
ReplyDeleteHi Elizabeth, I totally agree with you. I know estrogen does have an impact on adipose tissue formation and id assume that’s why they included menopausal status within the study. Further, I totally agree that news media has a bias to write the most impactful article and not the most transparent in most cases. At the end of the day their job is get clicks.
DeleteHi Rahib, well done on your blog post! Women-specific health studies can be sorely lacking in the overall literature, so it's nice to see research done with women in their midlife as the focus, especially seeing that it's not just white women. I was very impressed by the sheer number of factors the study accounted for and modeled against, as seen in Table 1, to show the correlation of air pollution with adipose. As others have mentioned, I'm very interested to see how the statistics may change with data from other regions. In regards to the news article, I agree that they cherry picked some data to appear more shocking and click-baity, but I don't think the inclusion of other experts necessarily undermines the primary researchers, seeing as the connection between exercise and gaining adipose tissue is quite well known.
ReplyDeleteI totally agree that women’s health is field that often gets neglected. I thought it was nice to see that as well. Getting information from other regions hopefully wouldn’t be hard maybe modeling upwelling and downwelling of currents could help or even getting samples over another extended period.
DeleteGreat Job Rahib! This is one of the most interesting studies I've seen in a long time. I was very apprehensive when I first read the title because it seemed very hard to prove to me. After I saw the article, it seemed to have a lot of confounding variables and they needed to exclude a lot of data. This led me to believe the processing and interpretation of the data is very subjective, but I am very curious of what you think. Is the journal article accurate and trustworthy in your opinion? I also agree with your rating of the news article because I noticed it does generalize obesity and exercise at the end.
ReplyDeleteI thought that too when I first saw the article, but after reading about the biochemical effects that were mentioned in the paper how pollutants can inflame adipose tissue or create hormonal imbalances. I thought back to how pollutants affect male fertility, this made me feel like it’s not so far fetched after all. In my opinion, I do think it’s trustworthy because of connection to biochemistry. It would be nice to see this done to rats in a more isolated setting to get some data that isn’t just longitudinal and doesn’t have to deal with as many confounding variables.
Delete